prismadata 0.4.2__tar.gz → 0.4.4__tar.gz

{prismadata-0.4.2 → prismadata-0.4.4}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: prismadata
-Version: 0.4.2
+Version: 0.4.4
 Summary: Python client for the PrismaData location intelligence API
 License: MIT
 License-File: LICENSE

{prismadata-0.4.2 → prismadata-0.4.4}/prismadata/_http.py

@@ -103,6 +103,30 @@ class HttpClient:
     def close(self) -> None:
         self._client.close()
 
+    def create_worker_client(self) -> HttpClient:
+        """Create an independent HttpClient for use in a worker thread.
+
+        Shares the same auth but has its own httpx.Client, throttle state,
+        and rate limit tracking. This avoids thread-safety issues with the
+        shared httpx.Client and prevents throttle contention between workers.
+        """
+        worker = HttpClient.__new__(HttpClient)
+        worker._auth = self._auth
+        worker._timeout = self._timeout
+        worker._last_request_time = 0.0
+        worker._rl_remaining = None
+        worker._rl_reset = None
+        worker._quota_remaining = None
+        worker._quota_reset = None
+        worker._quota_limit = None
+        worker._quota_used = None
+        worker._quota_period = None
+        worker._client = httpx.Client(
+            timeout=self._timeout,
+            headers=dict(self._client.headers),
+        )
+        return worker
+
     def get(self, path: str, params: dict[str, Any] | None = None) -> Any:
         return self._request("GET", path, params=params)
 

{prismadata-0.4.2 → prismadata-0.4.4}/prismadata/async_client.py

@@ -661,7 +661,7 @@ class AsyncClient:
         timeout: int | None = None,
         show_progress: bool | None = None,
         auto_scale: bool = True,
-        max_workers: int = DEFAULT_MAX_WORKERS,
+        max_workers: int | None = None,
         chunk_threshold: int = DEFAULT_CHUNK_THRESHOLD,
         total_items_estimate: int | None = None,
         **kwargs: Any,
@@ -698,7 +698,8 @@ class AsyncClient:
             prepare_total = total_items_estimate if total_items_estimate is not None else total
             resp = await async_batch_prepare(self._post, prepare_total)
             session_id = resp["session_id"]
-            num_workers = min(max_workers, resp.get("max_workers", max_workers))
+            server_workers = resp.get("max_workers", DEFAULT_MAX_WORKERS)
+            num_workers = min(max_workers, server_workers) if max_workers is not None else server_workers
 
             try:
                 await async_wait_until_ready(self._get, session_id)

{prismadata-0.4.2 → prismadata-0.4.4}/prismadata/client.py

@@ -24,7 +24,7 @@ from ._constants import (
     MAX_BATCH_SIZE,
     MAX_ROUTING_BATCH,
 )
-from ._http import HttpClient
+from ._http import HttpClient, _handle_response
 from ._prepare import batch_complete, batch_prepare, wait_until_ready
 from ._progress import progress_bar
 from ._validation import validate_lat_lng, validate_profile, validate_route_points
@@ -888,7 +888,7 @@ class Client:
         timeout: int | None = None,
         show_progress: bool | None = None,
         auto_scale: bool = True,
-        max_workers: int = DEFAULT_MAX_WORKERS,
+        max_workers: int | None = None,
         chunk_threshold: int = DEFAULT_CHUNK_THRESHOLD,
         total_items_estimate: int | None = None,
         **kwargs: Any,
@@ -907,7 +907,8 @@ class Client:
             timeout: Override default request timeout (seconds).
             show_progress: Override progress bar setting.
             auto_scale: If True, call prepare/complete for large batches.
-            max_workers: Max parallel workers (server may return fewer).
+            max_workers: Max parallel workers. When auto_scale is True and
+                not set, the server decides. When set, caps the server value.
             chunk_threshold: Minimum items to trigger auto-scaling.
             total_items_estimate: Global job size estimate for infrastructure
                 sizing. If not provided, uses ``len(addresses)``.
@@ -937,24 +938,37 @@ class Client:
             prepare_total = total_items_estimate if total_items_estimate is not None else total
             resp = batch_prepare(self._post, prepare_total)
             session_id = resp["session_id"]
-            num_workers = min(max_workers, resp.get("max_workers", max_workers))
+            server_workers = resp.get("max_workers", DEFAULT_MAX_WORKERS)
+            num_workers = min(max_workers, server_workers) if max_workers is not None else server_workers
 
+            worker_clients: list[Any] = []
             try:
                 wait_until_ready(self._get, session_id)
                 groups = split_into_groups(addresses, num_workers)
 
                 from concurrent.futures import ThreadPoolExecutor, as_completed
 
-                def _process_group(group: dict) -> dict[str, Any]:
-                    return process_batch(group, _request, chunk_size, on_error=on_error)
+                worker_clients = [self._http.create_worker_client() for _ in range(num_workers)]
+
+                def _process_group(group: dict, worker_http: Any) -> dict[str, Any]:
+                    def _worker_request(chunk: dict) -> dict[str, Any]:
+                        return _handle_response(
+                            worker_http._do_request("POST", "/location/batch/geocoder/aggregator", params=params, json_body=chunk, timeout=timeout)
+                        )
+                    return process_batch(group, _worker_request, chunk_size, on_error=on_error)
 
                 with ThreadPoolExecutor(max_workers=num_workers) as executor:
-                    futures = [executor.submit(_process_group, g) for g in groups]
+                    futures = [
+                        executor.submit(_process_group, g, wc)
+                        for g, wc in zip(groups, worker_clients)
+                    ]
                     result = {}
                     for f in as_completed(futures):
                         chunk_result = f.result()
                         result.update(chunk_result)
             finally:
+                for wc in worker_clients:
+                    wc.close()
                 batch_complete(self._post, session_id)
 
         if self._clean:

{prismadata-0.4.2 → prismadata-0.4.4}/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "prismadata"
-version = "0.4.2"
+version = "0.4.4"
 description = "Python client for the PrismaData location intelligence API"
 authors = ["PrismaData <contato@prismadata.io>"]
 license = "MIT"